Using Iterative Strategies for Aligning Attributed Points in Space

Abstract

Signatures based on shape features are expected to be useful for several similarity assessment applications. Such signatures can extract the relevant shape information needed for similarity assessment and ignore the shape information not needed for a particular application. Since different feature-based models are created in different coordinate systems, in order to compute distance between two sets of features, one needs to include geometric transformations for aligning features. Optimal alignment is the alignment that results in the minimum distance between two sets of features. In general, features can be viewed as attributed points in space. Hence, we need to develop techniques for aligning attributed points in space. This report introduces iterative strategies for optimally aligning attributed points in space. Iterative strategies presented in this report involve successively applying algorithms that perform alignment under restricted rigid body transformations (e.g., rotations only or translations only) to solve point alignment problems that require higher dimension transformations (e.g., combinations rotations and translations). Our preliminary experimental results show that the idea of using of iterative strategies to solve higher dimension attributed point alignment problems is promising and can be used to perform feature-based shape similarity assessment.